Non-parametric methods have become essential in modern statistical theory and practice, offering flexibility and robustness in analyzing complex data structures without assuming a specific parametric form for the underlying distributions. The workshop will bring together specialists in the field to discuss recent advances in non-parametric statistics and related topics such as inverse problems, optimization, stochastic processes, and others.
The workshop will also honor Professor Vladimir Spokoiny's 65th birthday and his significant influence on modern nonparametrics.

Keynote speakers

• Denis Belomestny (Universiät Duisburg-Essen)
• Natalia Bochkina (University of Edinburgh)
• Alexander Goldenschluger (University of Haifa)
• Oleg Lepsky (Aix-Marseille Université)
• Enno Mammen (Universitä Heidelberg)
• Alexandre Tsybakov (CREST-ENSAE Paris)

Organizers

• Christian Bayer (WIAS Berlin)
• Pavel Dvurechensky (WIAS Berlin)
• Peter K. Friz (TU and WIAS Berlin)
• Markus Reiß (HU Berlin and Member of Steering Committee "SFB 1294 Data Assimilation")

Program

09:30 - 10:00	Registration
10:00 - 10:10	Opening
10:10 - 11:00	Oleg Lepski (Aix-Marseille Université ) Adaptive estimation of $L_2$-norm of a probability density and related topics
11:00 - 11:30	Coffee break
11:30 - 12:20	Denis Belomestny (Universiät Duisburg-Essen) On statistical complexity of the inverse optimal control
12:20 - 13:10	Alexandre Tsybakov (CREST-ENSAE Paris) A conversion theorem and minimax optimality for continuum contextual bandits
13:10 - 14:00	Lunch (catering at the venue)
14:00 - 14:50	Alexander Goldenshluger (University of Haifa) Statistics of Smoluchowski processes
14:50 - 15:40	Natalia Bochkina (University of Edinburgh) Bernstein - von Mises theorem in inverse problems
15:40 - 16:10	Coffee break
16:10 - 17:00	Enno Mammen (Universität Heidelberg ) Estimation of panel models with group structures in fixed effect In this talk we discuss panel models with unobserved individual effects. In this model parameters on time-varying covariates are identifiable and are consistently estimated by the classical fixed effects estimator. However parameters on the time-constant covariates are not identifiable. In this talk we present a new approach to clustering in this model to ensure identifiability. By using unsupervised nonparametric density-based clustering, cluster patterns including their location and number are adaptively determined. The approach works with large data structures. Our approach differs in two respects from the related literature. We allow for atoms, i.e. for units not belonging to a cluster and in our theoretical study we consider an asymptotic framework where the clusters are not consistently estimated in the limit. The performance of our method for large data sets is illustrated by simulations and an application to labour market data with 77,500 individuals and 620,000 person-year observations. The talk reports on joint work with Ralf A. Wilke {Copenhagen) and Kristina Zapp (Mannheim).

Venue

The workshop takes place at Humboldt University, Erwin-Schrödinger-Zentrum, Rudower Chaussee 26, 12489 Berlin - Adlershof. Venue on Google Maps.

Contact and further information

Registration is free but mandatory, via the online registration form. Deadline is October 1st.

If you have any questions, please do not hesitate to contact the organisers at: Spokoiny2024@wias-berlin.de

Accessibility, and day care for children

The venue is wheel-chair accessible by appointment. Kindly inform us about any particular needs. We can provide and sponsor child care during the workshop. Please do not hesitate to contact us.

Support

Weierstrass Institute for Applied Analysis and Stochastics, Leibniz Institute in Forschungsverbund Berlin e. V.

Humbold-Universität zu Berlin

Deutsche Forschungsgemeinschaft (DFG) via Individual Grants Program and Collaborative Research Center/Transregio "Rough Analysis, Stochastic Dynamics and Related Fields" (SFB/Transregio 388)

SFB 1294 - Data Assimilation